25-10-2016, 03:03 PM
1461133544-PHYSICS.docx (Size: 156.68 KB / Downloads: 2)
To provide a context for vehicle speed estimation algorithm, two methods-camera model and analytical expressions for features that can be measured in the images in terms of the camera parameters. Thegeometry of our camera and roadway are shownin Fig. 2. We model the scene as a set of parallel lines viewed through a pinhole camera, assuming that the camera is located at a height h above the ground plane and a perpendicular distance d from the edge of the roadway. The camera is oriented at a pan angle ѳ with respect to the road and tilt (down) angle Ҩ such that a point in the earth system (X,Y,Z) is transformed into the camera coordinate system by only a translation and a rotation. The camera is oriented along the negative axis (hence the negative sign associated with the focal length) and its line of sight intersects the ground plane a distance awayF=cosec(Ҩ)
This model is accurate if we assume the road is planar and the rotation of the ground plane about the axis is negligible. Based on our analysis with a full (three-angle) camera model, there is theoretically less than a 10% bias in distance measure- ments for road grades of 2% or less when using the reduced (two-angle) camera model. The following equations describe the perspective projection of points in the pinhole camera’s co- ordinate system onto the image, given a focal length :